merged 2.2.4

CMakeLists.txt

@@ -72,7 +72,7 @@ endif()
 
 set(PROJECT_NAME "oneDNN")
 set(PROJECT_FULL_NAME "oneAPI Deep Neural Network Library (oneDNN)")
-set(PROJECT_VERSION "2.2.3")
+set(PROJECT_VERSION "2.2.4")
 
 set(LIB_NAME dnnl)
 

src/common/primitive_cache.hpp

@@ -20,6 +20,7 @@
 #include <future>
 #include <list>
 #include <memory>
+#include <thread>
 #include <unordered_map>
 
 #include "c_types_map.hpp"

src/gpu/jit/ngen/ngen_auto_swsb.hpp

@@ -33,6 +33,7 @@
 
 #include <list>
 #include <map>
+#include <limits>
 
 namespace ngen {
 namespace autoswsb {

tests/benchdnn/inputs/conv/perf_conv_gen12lp

@@ -1,13 +0,0 @@
---batch=perf_conv_gen9
-
---reset
-
-# Forward, int8
---cfg=u8s8s8
---dir=FWD_B
-
---mb=16,64
---batch=set_perf_gpu_large_mb
-
---mb=1
---batch=set_perf_gpu_all_mb

tests/benchdnn/inputs/ip/perf_ip_gen12lp

@@ -1,13 +0,0 @@
---batch=perf_ip_gen9
-
---reset
-
-# Forward, int8
---dir=FWD_B
---cfg=u8s8s8
-
---mb=1,32,64,128
---batch=set_gpu
-
---mb=0
---batch=shapes_1d

tests/benchdnn/inputs/rnn/perf_rnn_gen12lp

@@ -1,13 +0,0 @@
---batch=perf_rnn_gen9
-
---reset
-
-# Forward, int8
---prop=FWD_D
---cfg=u8u8u8u8
---scaling=per_oc
---trivial-strides=true
-
---batch=option_set_perf_inference_lb
---batch=option_set_perf_inference_sb
---batch=option_set_perf_training

tests/gtests/in/gemm_in.h

@@ -121,30 +121,6 @@ CPU_INST_TEST_CASE(TestGEMM_smalln,
         test_params {'n', 't', 8, 512, 2048, 1.0f, 1.0f, 2048, 2048, 512},
         test_params {'n', 't', 8, 2048, 512, 1.0f, 1.0f, 512, 512, 2048});
 
-/**
- * These cases are used to test the small-N avx-512 sgemm TN kernels.
- * Note: The kernels assume a column major layout while the external 
- * APIs assume row major layout, so the M/N and transA/transB values 
- * are swapped.
- */
-CPU_INST_TEST_CASE(TestGEMM_smalln,
-        test_params {'n', 't', 5, 512, 512, 1.0f, 1.0f, 512, 512, 512},
-        test_params {'n', 't', 5, 512, 1536, 1.0f, 1.0f, 1536, 1536, 512},
-        test_params {'n', 't', 5, 512, 2048, 1.0f, 1.0f, 2048, 2048, 512},
-        test_params {'n', 't', 5, 2048, 512, 1.0f, 1.0f, 512, 512, 2048},
-        test_params {'n', 't', 7, 512, 512, 0.0f, 1.0f, 512, 512, 512},
-        test_params {'n', 't', 7, 512, 1536, 1.0f, 0.0f, 1536, 1536, 512},
-        test_params {'n', 't', 7, 512, 2048, 0.5f, 0.5f, 2048, 2048, 512},
-        test_params {'n', 't', 7, 2048, 512, 1.0f, 1.0f, 512, 512, 2048},
-        test_params {'n', 't', 4, 512, 512, 1.0f, 1.0f, 512, 512, 512},
-        test_params {'n', 't', 4, 512, 1536, 1.0f, 1.0f, 1536, 1536, 512},
-        test_params {'n', 't', 4, 512, 2048, 1.0f, 1.0f, 2048, 2048, 512},
-        test_params {'n', 't', 4, 2048, 512, 1.0f, 1.0f, 512, 512, 2048},
-        test_params {'n', 't', 8, 512, 512, 1.0f, 1.0f, 512, 512, 512},
-        test_params {'n', 't', 8, 512, 1536, 1.0f, 1.0f, 1536, 1536, 512},
-        test_params {'n', 't', 8, 512, 2048, 1.0f, 1.0f, 2048, 2048, 512},
-        test_params {'n', 't', 8, 2048, 512, 1.0f, 1.0f, 512, 512, 2048});
-
 #if defined(FP32) || defined(BF16BF16F32)
 INST_TEST_CASE(TestGEMM_packed,
         test_params {'t', 'n', 3, 2, 1, 1.0, 0.0, 2, 5, 8, {}, {false, true},